Content uploaded by Michael R Duchen
Author content
All content in this area was uploaded by Michael R Duchen
Content may be subject to copyright.
Changes in Intracellular Calcium and Glutathione in
Astrocytes as the Primary Mechanism of
Amyloid Neurotoxicity
Andrey Y. Abramov,
1
Laura Canevari,
2
and Michael R. Duchen
1
1
Mitochondrial Biology Group, Department of Physiology, University College London, London WC1E 6BT, United Kingdom, and
2
Miriam Marks Division
of Neurochemistry, Institute of Neurology, London WC1N 3BG, United Kingdom
Although the accumulation of the neurotoxic peptide

amyloid (

A) in the CNS is a hallmark of Alzheimer’s disease, the mechanism of

A neurotoxicity remains controversial. In cultures of mixed neurons and astrocytes, we found that both the full-length peptide

A
(1– 42) and the neurotoxic fragment (25–35) caused sporadic cytoplasmic calcium [intracellular calcium ([Ca
2⫹
]
c
)] signals in astrocytes
that continued for hours, whereas adjacent neurons were completely unaffected. Nevertheless, after 24 hr, although astrocyte cell death
was marginally increased, ⬃50% of the neurons had died. The [Ca
2⫹
]
c
signal was entirely dependent on Ca
2⫹
influx and was blocked by
zinc and by clioquinol, a heavy-metal chelator that is neuroprotective in models of Alzheimer’s disease. Neuronal death was associated
with Ca
2⫹
-dependent glutathione depletion in both astrocytes and neurons. Thus, astrocytes appear to be the primary target of

A,
whereas the neurotoxicity reflects the neuronal dependence on astrocytes for antioxidant support.
Key words:

-amyloid; intracellular calcium; astrocyte; neuron; Alzheimer; glutathione
Introduction
Alzheimer’s disease (AD) is a neurodegenerative disorder char-
acterized by a progressive cognitive decline resulting from selec-
tive neuronal dysfunction, synaptic loss, and neuronal cell death.
It is accompanied by the deposition of the

-amyloid peptide
(

A), a polypeptide of 39 – 43 aa that is thought to play a major
role in the pathogenesis of the disorder (Small and McLean,
1999). Aggregated

A is neurotoxic. Although

A neurotoxicity
has been associated with oxidative stress and the reduction of
endogenous antioxidants (Behl et al., 1994; Casley et al., 2002a),
with mitochondrial damage (Casley et al., 2002b) and with the
destabilization of intracellular calcium ([Ca
2⫹
]
c
) homeostasis,
both in neurons (Mattson et al., 1992) and in glial cells (Stix and
Reiser,1998), the mechanism of

A-induced neurotoxicity re-
mains uncertain.
Amyloid

peptides have been shown recently to form pores
in artificial membranes; it has been suggested that they may also
act as pore formers in intact neuronal membranes, in which they
appear to form Ca
2⫹
-permeable channels (Arispe et al., 1993; Lin
et al., 2001).

A has also been shown to have effects on a variety of
types of ion-selective channels, including voltage-gated Ca
2⫹
-
permeant channels (Blanchard et al., 1997; Ueda et al., 1997;
Rovira et al., 2002). More subtle changes in [Ca
2⫹
]
c
signaling
have also been demonstrated after long-term exposure to

A
(Mattson and Chan, 2001), suggesting a disturbance of [Ca
2⫹
]
c
homeostatic mechanisms that may reflect changes in cellular
metabolism.
Reports on the effects of

Aon[Ca
2⫹
]
c
in astrocytes are con-
troversial. With exposure to

A, some authors have found that
astrocyte [Ca
2⫹
]
c
increases (Stix and Reiser, 1998), whereas in
the hands of others, astrocyte [Ca
2⫹
]
c
decreases (Meske et al.,
1998). The effect of

Aon[Ca
2⫹
]
c
homeostasis in astrocytes is
potentially very important, considering the interplay between
neuronal and glial signals revealed recently (Haydon, 2001) and
the proposed glia-related pathomechanisms in AD (Harkany et
al., 2000; Schubert et al., 2001). Glial cells play a major supportive
role toward neurons, which includes supplying metabolic sub-
strates and the precursors of the antioxidant glutathione (GSH)
(Dringen, 2000) and removing excitatory amino acids such as gluta-
mate from the extracellular space (Takahashi et al., 1997), processes
that play a critical role in neuroprotection. These roles may be un-
dermined by the

A-induced generation of reactive oxygen species
(ROS) and the inhibition of glutamate uptake (Markesbery,1997;
Harkany et al.,2000), resulting in neuronal damage as a consequence
of impaired astrocytic support function.
Materials and Methods
Cell culture. Mixed cultures of hippocampal neurons and glial cells were
prepared as described previously (Vergun et al., 2001), with modifica-
tions, from Sprague Dawley rat pups 2– 4 d postpartum [University Col-
lege London (UCL) breeding colony]. Hippocampi were removed into
ice-cold Gey’s salt solution (Invitrogen, Paisley, UK) with 20
g/ml of
gentamicin. The tissue was minced and trypsinized (0.1% for 15 min at
37°C), triturated, and plated on poly-D-lysine-coated coverslips and cul-
tured in Neurobasal medium (Invitrogen) supplemented with B
27
(In-
vitrogen) and 2 mML-glutamine. Cultures were maintained at 37°C in a
humidified atmosphere of 5% CO
2
and 95% air, fed twice a week, and
maintained for a minimum of 10 d before experimental use to ensure the
expression of glutamate and other receptors. Neurons were easily distin-
guishable from glia: they appeared phase-bright, had smooth, rounded so-
Received Feb. 12, 2003; revised April 4, 2003; accepted April 10, 2003.
This work was supported by the Wellcome Trust, the Royal Society, and the Miriam Marks Foundation. We thank
Drs. Frances Edwards and Anna de Simoni for providing hippocampal explant cultures and Profs. S. Bolsover and J. B.
Clark and Drs. R. Dumollard and J. Jacobson for their invaluable discussion and suggestions.
Correspondence should be addressed to Michael R. Duchen, Department of Physiology, University College Lon-
don, Gower Street, London WC1E 6BT, UK. E-mail: m.duchen@ucl.ac.uk.
Copyright © 2003 Society for Neuroscience 0270-6474/03/235088-08$15.00/0
5088 •The Journal of Neuroscience, June 15, 2003 •23(12):5088 –5095
mata and distinct processes, and lay just above the focal plane of the glial
layer. Cells were used at 10–20din vitro (DIV) unless stated otherwise.
Isolated cortical astrocytes were prepared as described previously
(Boitier et al., 1999). Cerebra taken from adult Sprague Dawley rats (UCL
breeding colony) were chopped and triturated until homogeneous,
passed through a 297
m mesh, and trypsinized (50,000 U/ml of porcine
pancreas; Sigma, Gillingham, UK) with 336 U/ml of DNase 1 (bovine
pancreas, Sigma), and 1.033 U/ml of collagenase (Sigma) at 37°C for 15
min. After the addition of fetal bovine serum (10% of final volume) and
filtering through 140
Mmesh, the tissue was centrifuged through 0.4 M
sucrose (400 gm, 10 min), and the resulting pellet was transferred to
Minimal Essential Medium supplemented with 5% fetal bovine serum, 2
mMglutamine, and 1 mMmalate in tissue culture flasks precoated with
0.01% poly-D-lysine. The cells reached confluence at 12–14 DIV; they
were harvested and reseeded onto 24 mm diameter glass coverslips (BDH
Chemicals, Poole, UK) precoated with 0.01% poly-D-lysine for fluores-
cence measurements, and used over 2–4d.
Peptides and treatments.

A25–35,

A1–42, and

A35–25 (Bachem, St.
Helens, UK) were dissolved at 1 mMin sterile ultrapure water (Milli-Q stan-
dard; Millipore, Watford, UK) and kept frozen until use. The peptides were
added under the microscope, except for GSH and neurotoxicity measure-
ments, where they were added 24 hr before the experiment.

A25–35 was
used at concentrations of up to 50
Mto ensure that it was present in molar
excess compared with inhibitors and so to exclude any direct interaction.
Microscopy. Fluorescence measurements were obtained using a Nikon
(Tokyo, Japan) epifluorescence inverted microscope with a 20⫻fluorite
objective. Excitation light from a Xenon arc lamp is selected using 10 nm
bandpass filters centered at 340, 360, and 380 nm housed in a computer-
controlled filter wheel (Cairn Research, Faversham, UK). Emitted light
passed through a long-pass filter to a cooled CCD camera (Orca ER;
Hamamatsu, Welwyn Garden City, UK). All imaging data were collected
at intervals of 10–15 sec, digitized, and analyzed using Kinetic Imaging
(Wirral, UK) software. Cells were protected from phototoxicity by inter-
posing a shutter in the light path to limit exposure between the acquisi-
tion of successive images.
Confocal images were obtained using a Zeiss (Oberkochen, Germany)
510 confocal laser scanning microscope and a 40⫻oil immersion objec-
tive. The 488 nm argon laser line was used to excite fluo-4 fluorescence,
which was measured using a bandpass filter from 505 to 550 nm. Illumi-
nation intensity was kept to a minimum (at 0.1% of laser output) to avoid
phototoxicity, and the pinhole was set to give an optical slice of ⬃2
m.
[Ca
2⫹
]
c
measurements. Cells were loaded for 30 min at room temper-
ature with 5
Mfura-2 AM (Molecular Probes, Eugene, OR) and 0.005%
Pluronic in a HEPES-buffered salt solution composed of (in mM): 156
NaCl, 3 KCl, 2 MgSO
4
, 1.25 KH
2
PO
4
, 2 CaCl
2
, 10 glucose, and 10 HEPES,
pH 7.35. Traces, obtained using the cooled CCD imaging system, are
presented as ratios of excitation at 340 and 380 nm, both with emission at
⬎515 nm. For some measurements, [Ca
2⫹
]
i
was calculated using the
equation (Grynkiewicz et al., 1985): [Ca
2⫹
]
c
⫽K(R⫺R
min
)/(R
max
⫺R),
where Ris the fluorescence ratio (340/380 nm) and Kis the effective
dissociation constant of fura-2. R
max
and R
min
were determined by the
application of 50
Mdigitonin followed by 1 mMMnCl
2
.
All data presented were obtained from at least five coverslips and two
to three different cell preparations.
For confocal imaging, cells were loaded with fluo-4 AM (5
M; Molec-
ular Probes) for 20 min, followed by washing. Data are presented nor-
malized with respect to the first image of the sequence.
GSH measurements. To measure GSH, cells were incubated with 50
M
monochlorobimane (MCB; Molecular Probes) in HEPES-buffered salt
solution at room temperature for 40 min, or until a steady state had been
reached before images were acquired for quantitation (Keelan et al.,
2001). The cells were then washed with HEPES-buffered salt solution,
and images of the fluorescence of the MCB-GSH adduct were acquired
using the cooled CCD imaging system as described using excitation at
380 nm and emission at ⬎400 nm.
Toxicity experiments. For toxicity assays we loaded cells simultaneously
with 20
Mpropidium iodide (PI), which is excluded from viable cells
but exhibits a red fluorescence after a loss of membrane integrity, and 4.5
MHoechst 33342 (Molecular Probes), which gives a blue stain to chro-
matin, to count the total number of cells. Using phase-contrast optics, a
bright-field image allowed the identification of neurons, which look
quite different from the flatter glial component and also lie in a different
focal plane, above the glial layer. A total of 600 –800 neurons or glial cells
were counted in 20 –25 fields of each coverslip. Each experiment was
repeated five or more times using separate cultures.
Statistical analysis. Statistical analysis and exponential curve fitting
were performed using Origin 7 (Microcal Software Inc., Northampton,
MA) software. Results are expressed as means ⫾SEM.
Results
We have used digital imaging techniques to explore the effects of

Aon[Ca
2⫹
]
c
signals, antioxidant status, and cell viability in
cultures of mixed hippocampal neurons and astrocytes in cul-
tures ranging from 10 to 45 DIV. Application of either the full-
length peptide (1–42; 100 nMto 10
M) or the 25–35 aa fragment
(1–50
M)of

A had no effect on [Ca
2⫹
]
c
signals in neurons,
which remained quiescent over periods of up to 6 hr (n⫽456
cells) (Fig. 1Aa,C). After 30–40 min of incubation with

A the
cells showed robust and reversible responses to the application of
Figure 1.

Amyloid raises [Ca
2⫹
]
c
in astrocytes and not in neurons. A, Records of fura-2
fluorescence from neurons ( a) and astrocytes (b) in hippocampal cocultures after exposure to

A25–35 peptide (50
M). The neurons showed no change in signal over a period of 35 min.
Their identity was confirmed by their response to glutamate (100
M) at the end of this period.
The astrocytes ( b) showed complex [Ca
2⫹
]
c
fluctuations starting after ⬃5– 6 min of exposure
to

A25–35 (50
M). These could continue for many hours. Some sample traces are extracted
from this population and illustrated as Bi,Bii, and Biii. The images in Care taken from a time
series of confocal images of a hippocampal coculture loaded with fluo-4. The field includes four
neurons (n) surrounded by astrocytes. Once again, the astrocytes show complex transient and
localized [Ca
2⫹
]
c
responses whereas the neurons show no change in signal at all. The traces in
Doriginate from confocal images of a fluo-4-loaded hippocampal explant culture. Once again,
neuronsthatshowed a robust responsetoglutamate application showed no responseto

A(a)
whereas astrocytes showed complex [Ca
2⫹
]
c
fluctuations and only a small transient metabo-
tropic [Ca
2⫹
]
c
response to glutamate ( b). An image of a responding cell is inset for each signal.
Abramov et al. •Astrocytic Calcium and GSH Changes in Amyloid Neurotoxicity J. Neurosci., June 15, 2003 •23(12):5088 –5095 • 5089
glutamate (100
M) (Fig. 1Aa) or to depolarization with 50 mM
KCl, confirming both their viability and their neuronal identity.
Moreover, the presence of

A25–35 or

A1–42 did not change
the amplitude of the [Ca
2⫹
]
c
response to glutamate or to 50 mM
KCl; the response to the latter was 726 ⫾89 nM(n⫽59 cells) in
control and 759 ⫾97 nM(n⫽96 cells) in

A-incubated neurons.
Remarkably, astrocytes in the same cultures showed dramatic
[Ca
2⫹
]
c
signals after exposure to

A, often occurring in cells
surrounding quiescent neurons (Fig. 1Ab,B,C) Similar [Ca
2⫹
]
c
signals were seen in response to both the 25–35 peptide and the
full-length 1–42 peptide, but no responses were seen to the re-
verse peptide 35–25 (n⫽135 cells). All [Ca
2⫹
]
c
signals started
after a delay of ⬃5–15 min and showed three patterns of re-
sponse: (1) sporadic increases in [Ca
2⫹
]
c
, seen as low-amplitude
(100–200 nM) [Ca
2⫹
]
c
oscillations or fluctuations; (2) larger
spikes, followed by sustained elevated [Ca
2⫹
]
c
(1–2
M), and (3)
very large increases in [Ca
2⫹
]
c
, usually followed by loss of cell
viability. After washing the cells with

A-free saline, the re-
sponses persisted for up to 6 hr (data not shown).
Because responses like these have not been described previ-
ously, we were concerned that the properties of the cells might be
dictated by our culture conditions. Therefore, we repeated the
experiments using other culture systems. Experiments using cul-
tures of cortical astrocytes prepared in the same way gave results
identical to those from the hippocampus (data not shown). We
also used hippocampal explant cultures, in which the properties
of the tissue in vivo are well retained. Confocal imaging of explant
cultures loaded with fluo-4 again showed that exposure to

A25–35 (50
M) provoked a selective increase in activity in glial
cells in the culture (Fig. 1 Db). Neurons were identified as the only
cells in the culture to show a rise in [Ca
2⫹
]
c
with 50 mMKCl and
by their larger and more sustained response to glutamate (100
M) (Fig. 1Da). The astrocytes showed transient and oscillatory
activity with

A, showed no response to 50 mMKCl, and their
response to glutamate was a small transient response that reflects
activation of metabotropic glutamate receptors (n⫽5 cultures).
Because the responses in dissociated cultures and explants ap-
peared similar, the dissociated cultures were used for the remain-
der of the experiments described.
Astrocyte [Ca
2ⴙ
]
c
responses to

A are dependent on
extracellular Ca
2ⴙ
and independent of intracellular
Ca
2ⴙ
stores
The [Ca
2⫹
]
c
responses to

A were never observed in Ca
2⫹
-free
saline (n⫽231 cells) (Fig. 2A). However, if cells were exposed to

AinaCa
2⫹
-free saline and were then washed with

A-free,
Ca
2⫹
-containing buffer, [Ca
2⫹
]
c
responses were then seen in the
astrocytes (n⫽69 cells) (Fig. 2B, bottom) whereas neuronal
[Ca
2⫹
]
c
did not change beyond the small increase associated with
the restoration of basal Ca
2⫹
influx (n⫽64 cells) (Fig. 2B, top).
These data show that (1) the changes in [Ca
2⫹
]
c
in astrocytes are
initiated through Ca
2⫹
influx from external sources, (2) the ini-
tiation of the action of

A does not require the presence of Ca
2⫹
,
and (3) the effect persists despite removal of

A from the saline.
The oscillatory [Ca
2⫹
]
c
signals appeared typical of IP
3
-
mediated [Ca
2⫹
]
c
release from endoplasmic reticulum (ER) seen
in astrocytes in response to a range of agonists (Peuchen et al.,
1996) and to mechanical stimulation (Charles et al., 1991).
Therefore, we considered whether external Ca
2⫹
acts as a trigger
to activate phospholipase C (PLC), which would generate IP
3
and
so mobilize ER Ca
2⫹
. However, the experiments illustrated in
Figure 3 suggest that PLC- and IP
3
-mediated signaling do not
play a significant role in the

A-induced [Ca
2⫹
]
c
signals. Thus,
U73122 (5
M), an inhibitor of phospholipase C (Fig. 3A) did not
significantly impair the [Ca
2⫹
]
c
astrocyte responses to

A(n⫽
89 cells). Similarly, 2-APB, (40
M) an inhibitor of IP
3
-
dependent Ca
2⫹
release, failed to reduce

A-induced [Ca
2⫹
]
c
signals in astrocytes (n⫽35 cells) (Fig. 3B), whereas it completely
blocked the [Ca
2⫹
]
c
increase induced by ATP (100
M; data not
shown), which acts at purinergic receptors (P
2U
) to promote IP
3
-
dependent ER Ca
2⫹
release (Peuchen et al., 1996).
Although the expression and role of ryanodine receptors
(RyRs) in astrocytes is debatable (Matyash et al., 2002), we also
tested the effect of the RyR inhibitor dantrolene (10
m), which
again had no significant effect on the

A-induced [Ca
2⫹
]
c
signal
(n⫽46 cells) (Fig. 3C). The incubation of astrocytes with 0.1–1
Mthapsigargin (an inhibitor of ER Ca
2⫹
pumps) completely
depleted Ca
2⫹
from the ER, demonstrated by the absence of a
[Ca
2⫹
]
c
response to ATP (100
M) (Fig. 3D). The addition of

A
again then induced a [Ca
2⫹
]
c
response that was not significantly
different from the control responses (Fig. 3D). In this instance,
values of peak [Ca
2⫹
]
c
after thapsigargin were 390 ⫾54 nM,
compared with control responses to

A with peak values of 456 ⫾
57 nM(n⫽301 cells; p⬎0.05). We also noted that the resting
Ca
2⫹
level, which was usually slightly elevated after exposure to
thapsigargin because of the activation of store-operated Ca
2⫹
influx, was slightly depressed by

A, suggesting that, if anything,
Figure 2. [Ca
2⫹
]
c
responses to

A are dependent on extracellular Ca
2⫹
.A, In the absence
of external Ca
2⫹
, cortical astrocytes showed no change in [Ca
2⫹
]
c
after exposure to

A25–35
(50
M). B, In a coculture exposed to

A25–35 (50
M) in the absence of external Ca
2⫹
,no
response was seen in either neurons or astrocytes.

A was then washed out and external Ca
2⫹
added. Despite the removal of the

A, the addition of Ca
2⫹
caused a large increase in [Ca
2⫹
]
c
in the astrocytes but only a very small change in the neurons, reflecting the restoration of basal
calcium entry. Once again, the neuronal identity was confirmed by the response to 100
M
glutamate at the end of the experiment.
5090 •J. Neurosci., June 15, 2003 •23(12):5088 –5095 Abramov et al. •Astrocytic Calcium and GSH Changes in Amyloid Neurotoxicity

A suppresses store-operated Ca
2⫹
influx. Taken together, these
data strongly suggest that ER-stored Ca
2⫹
does not play a signif-
icant role in

A-induced [Ca
2⫹
]
c
signals.
Mitochondria may have a high Ca
2⫹
content in astrocytes
(Boitier et al., 1999), and we cannot overlook the possibility of the
participation of mitochondria in

A-induced [Ca
2⫹
]
c
signals,
especially because A

causes mitochondrial damage (Casley et
al., 2002b). After the depolarization of mitochondria with the
uncoupler carbonyl cyanide p-trifluoromethoxyphenylhydra-
zone (FCCP; 1
M),

A25–35 or

A1–42 still increased [Ca
2⫹
]
c
(n⫽84 cells). In the presence of FCCP, the [Ca
2⫹
]
c
signals were
significantly increased (789 ⫾36 vs 456 ⫾58 nM;p⬍0.005). This
could result from a fall in ATP or through the loss of mitochon-
drial Ca
2⫹
uptake. Nevertheless, FCCP did not reduce the

A-
induced Ca
2⫹
signal, showing that mitochondria cannot repre-
sent a significant source of Ca
2⫹
.
Thus, taken together, these data strongly suggest that

A
causes [Ca
2⫹
]
c
signals in astrocytes, but not in neurons, by in-
ducing a pathway for Ca
2⫹
influx across the plasma membrane.

A induces Ca
2ⴙ
influx into astrocytes
Extracellular Mn
2⫹
enters cells via Ca
2⫹
-permeant channels and
quenches the fluorescence of intracellular fura-2. This is most
readily seen when the fura-2 is excited at ⬃360 nm, the Ca
2⫹
-
independent (isosbestic) point of the fura-2 excitation spectrum,
whereas the Ca
2⫹
-dependent change of the 340/380 nm ratio is
not altered. In the presence of

A(n⫽154 cells), the signal
excited at 360 nm was unaltered in the absence Mn
2⫹
, showing
that this reliably reports a Ca
2⫹
-independent signal (Fig. 4A).
However, in the presence of 40
MMn
2⫹
, the 360 nm fura-2
signal showed stepwise and irreversible decreases in the signal
corresponding with each transient increase in [Ca
2⫹
]
c
(two exam-
ples are shown in Fig. 4B)
.
This approach also allowed us to test
whether

A caused Ca
2⫹
entry in neurons that was masked by Ca
2⫹
buffering. However,

A had no effect on the 360 nm fura-2 signal in
neurons (n⫽120 cells) in the presence of Mn
2⫹
, or, as shown above,
on the fura-2 ratio, confirming the selectivity of the action of

Aon
astrocytes. These observations suggest that each [Ca
2⫹
]
c
transient
reflects a pulse of Ca
2⫹
influx into the astrocytes.
Additional confocal imaging experiments showed that the

A-induced [Ca
2⫹
]
c
signal was initiated as a rapid focal increase
in [Ca
2⫹
]
c
. These responses could sometimes be seen clearly
originating from a point source (Fig. 5A,B), followed by slower
diffusion into the cytosol (Fig. 5A). In the examples in Figure 5B,
the rise in [Ca
2⫹
]
c
was restricted to a small part of the cell and
failed to extend through the cytoplasm. This again is consistent
with the activation of an influx pathway followed by Ca
2⫹
buff-
ering rather than the mobilization of ER stores, in which the
amplitude and rate of rise of the signal are maintained by active
propagation (Boitier et al., 1999).
Routes for Ca
2ⴙ
influx
According to some authors (Ueda et al., 1997; He et al., 2002)

A
may induce a [Ca
2⫹
]
c
signal in neurons by increasing Ca
2⫹
influx
through voltage-dependent calcium channels (VDCC). Because as-
trocytes in our cultures do not show a [Ca
2⫹
]
c
response to 50 mM
KCl, it seems unlikely that they express VDCC. Nifedipine (1
M),
an inhibitor of L-type VDCCs, had no effect on the shape or ampli-
tude of the

A-induced [Ca
2⫹
]
c
signals (1–42 or 25–35) in either
cortical or hippocampal astrocytes (n⫽56 cells), but it completely
blocked the [Ca
2⫹
]
c
response to 150 mMKCl in hippocampal neu-
rons. The responses were also not significantly affected by inhibitors
of either ionotropic or metabotropic glutamate receptors, including
20
MCNQX (n⫽98 cells), 10
M(⫹)-5-methyl-10,11-dihydro-
5H-dibenzo [a,d] cyclohepten-5,10-imine maleate (n⫽69 cells) or
50
M(S)-(⫾)-amino-4-carboxy-methyl-phenylacetic acid (n⫽
178 cells) (data not shown), suggesting that the responses do not
reflect glutamate release into the culture.
An additional Ca
2⫹
influx pathway expressed by glial cells and
probably not in neurons is the pathway for capacitative influx
(capacitative Ca
2⫹
entry, CCE). One possibility is that

A acts
through altering the opening probability of this pathway. There-
fore, we tested the action of lanthanum, which blocks CCE (Pizzo
et al., 2001). However, La
3⫹
(1
M) had no effect on the re-
sponses (n⫽67 cells) (data not shown).
It has been suggested that in some cell types,

A-induced
increases in the generation of ROS serve as a trigger, which then
raise [Ca
2⫹
]
c
(Varadarajan et al., 2000). The incubation of corti-
cal and hippocampal astrocytes with the antioxidant trolox (750
M) and ascorbate (1 mM, 45 min preincubation; n⫽67 cells) or
the superoxide scavenger 4-hydroxy-2,2,6,6-tetramethyl-
piperadine-1-oxyl (500
M) plus catalase (250 U/ml) (n⫽69
cells), which we have shown previously to be effective scavengers
of ROS (Vergun et al., 2001), did not have any significant impact
on the [Ca
2⫹
]
c
response of the cells to

A (data not shown),
suggesting that the production of ROS by

A is not responsible
for the [Ca
2⫹
]
c
increases in astrocytes.
Figure 3. Intracellular Ca
2⫹
stores do not make a significant contribution to the astrocyte
[Ca
2⫹
]
c
response to

A. Manipulations that either block components of the IP
3
and ryanodine
signaling pathways or that empty ER stores do not significantly alter the astrocyte responses to

A.Theapplicationof 50
M

A25–35caused[Ca
2⫹
]
c
transientsincorticalastrocytes despite
thepresenceof5
MU73122(aninhibitorof PLC; A), 40
M2-APB ( B),dantrolene(an inhibitor
of ryanodine receptors; C), and 1
Mthapsigargin (an inhibitor of ER Ca
2⫹
pumps; D), at a
concentration that prevented the response to ATP (100
M).
Abramov et al. •Astrocytic Calcium and GSH Changes in Amyloid Neurotoxicity J. Neurosci., June 15, 2003 •23(12):5088 –5095 • 5091
Zinc and clioquinol abolish the

A-
induced [Ca
2ⴙ
]
c
response in astrocytes

A peptides have been shown to form chan-
nels in artificial and biological membranes
(Arispe et al., 1993; Lin et al., 2001; Kawa-
hara et al., 1997). Such channels can be
blocked by Zn
2⫹
(Arispe et al., 1996). We
found that the incubation of hippocampal
or cortical astrocytes (n⫽253 cells) with up
to1m
MZnCl
2
completely prevented the ef-
fect of

Aon[Ca
2⫹
]
c
(Fig. 6A). However,
the addition of Zn
2⫹
had no effect on the

A-induced [Ca
2⫹
]
c
signals once they had
already started (data not shown), suggesting
that Zn
2⫹
is not acting simply to block
channels, but rather to prevent their
formation.

A peptide binds to metal ions with a
selectivity Cu
2⫹
⬎Fe
3⫹
⬎Zn
2⫹
(At-
wood et al., 1998), all of which promote
aggregation. The addition of Cu
2⫹
(1
M
to1mM) did not change the amplitude
(456 ⫾58 to 490 ⫾56 nM;n⫽81 cells) or
shape of the [Ca
2⫹
]
c
signals in either cor-
tical or hippocampal astrocytes (Fig. 6B),
suggesting that endogenous heavy-metal
ions present in the culture are sufficient to
promote the aggregation of

A.
Cu
2⫹
can undergo redox cycling and
generate ROS, whereas Zn
2⫹
is not redox-
active but competes with Cu
2⫹
for binding,
and therefore inhibits the oxidant proper-
ties of metal-bound

A (Cuajungco et al.,
2000). The inhibition of [Ca
2⫹
]
c
signals by
Zn
2⫹
does not appear to be dependent on
these redox properties, because: (1) we see
identical effects with both the full-length
peptide and the 25–35 fragment, which does
not have the metal-binding coordination
site, and (2) we see no effect of antioxidants
on the responses (above). Trace amounts of
metals may promote aggregation in an
␣
-helix conformation, whereas a high con-
centration of Zn
2⫹
(and to a lesser extent,
Cu
2⫹
) promote

-sheet fibrillar aggrega-
tion, which is classically associated with

A
toxicity. The channels blocked by Zn
2⫹
have an
␣
-helical structure, whereas the
mechanism described here seems more
likely to involve

-sheet formation because
(1) Zn
2⫹
prevents but does not reverse the
response and (2)

A25–35 cannot form
␣
-helical channels but can form

-sheets.
Four types of conductances have been ob-
served with

A1–42 in lipid bilayers (Kou-
rie et al., 2001).
Clioquinol, a chelator of Cu
2⫹
,Zn
2⫹
,
and Fe
2⫹
, prevents aggregation and res-
olubilizes

A; it has also been shown to have a beneficial effect in
mouse models of AD (Cherny et al., 2001; Melov, 2002). The
preincubation of cells with 1–2
Mclioquinol for 30 min dramat-
ically prevented the effect of

Aon[Ca
2⫹
]
c
of cortical astrocytes
(n⫽207 cells) (Fig. 6C).

A25–35 and

A1– 42 deplete GSH in hippocampal neurons
and astrocytes
We then explored the consequences of

A exposure for GSH,
using fluorescence imaging of the indicator MCB to identify
changes in GSH in different cell types within the same culture
Figure 4. Mn
2⫹
quench confirms that astrocyte [Ca
2⫹
]
c
transients reflect transient Ca
2⫹
influx. Fura-2-loaded hippocampal
astrocytes showed typical [Ca
2⫹
]
c
fluctuations (black line) in response to 50
M

A25–35. A, In the absence of external Mn
2⫹
,
the fura-2 response excited at 360 nm (gray line) showed no change during the [Ca
2⫹
]
c
transients, confirming that this is close to
the isosbestic [Ca
2⫹
]
c
-independent excitation wavelength for fura-2. Bi, Bii, With the addition of 40
MMn
2⫹
each [Ca
2⫹
]
c
transient was accompanied by a step quench of the 360 nm fura-2 signal, confirming that each transient reflects a pulsed influx of
divalent cations seen in response to

A25–35.
Figure 5. Confocal imaging reveals focal Ca
2⫹
influx in response to

A. In a hippocampal coculture loaded with fluo-4,
confocalimagingduring the exposure to

Ashowsthat the change in[Ca
2⫹
]
c
canoriginateas a focal changethatdiffuses through
the cell and may be restricted to the subplasmalemmal space. Aa, Time series of confocal images taken during a single [Ca
2⫹
]
c
transient response in an astrocyte. Note that the response begins with a focal rise in [Ca
2⫹
]
c
(arrowhead) followed by the slower
spread through the cell. This is illustrated further in Ab, which shows a plot of the signal with time at four different locations in the
cell (indicated color-coded on the inset image). The rapid rate of rise at the point of influx contrasts with the much slower increase
seendeepinthe cytosol of the cell.B,Series of images taken fromanotherastrocyteduring a response to

A25–35,againshowing
that the [Ca
2⫹
]
c
signal may be restricted to the periphery of the cell and fail to propagate through the cell. The first image of the
sequenceshowsthe raw data, whereasthesubsequent images showtheratio of the imagesequencewith respect tothefirst image
of the sequence.
5092 •J. Neurosci., June 15, 2003 •23(12):5088 –5095 Abramov et al. •Astrocytic Calcium and GSH Changes in Amyloid Neurotoxicity
(Keelan et al., 2001). In agreement with previous reports (Casley
et al., 2002a; Muller et al., 1997; White et al., 1999), we found that

A significantly decreased GSH in cortical astrocyte monocul-
tures (by 54.7 ⫾4.9%; n⫽798 cells; p⬍0.001) (Fig. 7C) and in
hippocampal astrocytes in coculture (by 44.32 ⫾4.9%; control ⫽
100%) (Fig. 7A)(n⫽831 cells; p⬍0.005) after a 24 hr
incubation.
In contrast to the lack of effect of

A on neuronal [Ca
2⫹
]
c
,

A
also significantly ( p⬍0.01) reduced GSH in hippocampal neu-
rons (33.5 ⫾6.3%; n⫽345 cells) (Fig. 7B). Because astrocytes
supply neighboring neurons with GSH precursors (Sagara et al.,
1993) the GSH decline in the neurons is
likely to be a secondary consequence of
the decrease of astrocyte GSH.
The removal of external Ca
2⫹
alone
had no effect on GSH levels in hippocam-
pal neurons (n⫽420 cells) or astrocytes
(n⫽905 cells) in control experiments
(Fig. 7A–C). However, the effect of

A1–42 and

A25–35 on GSH was abol-
ished in the absence of Ca
2⫹
in both hip-
pocampal or cortical astrocytes and in
hippocampal neurons (Fig. 7A–C). Thus,
given the dependence of

A-induced
[Ca
2⫹
]
c
fluctuations on external Ca
2⫹
,it
seems likely that the GSH changes are a
direct consequence of the changes in as-
trocyte [Ca
2⫹
]
c
.
In the presence of 1–2
Mclioquinol, a
concentration that abolished

A-induced
[Ca
2⫹
]
c
fluctuations in astrocytes,

A25–35 (and

A1–42, data
not shown) no longer caused a significant fall of GSH in cortical
astrocytes (n⫽619 cells) (Fig. 7C).
Effect of

A on cell viability
We then examined the effect of a 24 hr exposure of cultures to

A25–35 on cell viability and found that, remarkably, 49.9 ⫾
8.5% of neurons (Fig. 8A) but only 23.2 ⫾4.2% of astrocytes
(n⫽9 experiments) (Fig. 8B) died during this period in hip-
pocampal cocultures. Preincubation with 1
Mclioquinol re-
duced cell death of hippocampal neurons and of cocultured hip-
pocampal astrocytes by ⬃50% (21.2 ⫾6.8% dead cells, p⬍0.05,
and 15.2 ⫾3.2%, respectively, n⫽5 experiments) (Fig. 8).
The removal of Ca
2⫹
from the medium also significantly ( p⬍
0.001) protected the hippocampal neurons (cell death fell from
49.9 ⫾8.5 to 16.45 ⫾4.1%; n⫽6 experiments) and in cocultured
hippocampal astrocytes (from 23.2 ⫾4.2 to 17.5%; n⫽7 exper-
iments; p⬎0.05). The presence or absence of Ca
2⫹
in the me-
dium did not change the percentage of dead cells in untreated
cells or in cells treated with the reverse peptide 35–25 (Fig. 8 A,B).
Discussion
We have found that

A induces calcium signals selectively in
astrocytes causing sporadic fluctuations of [Ca
2⫹
]
c
, while having
no apparent effect at all on [Ca
2⫹
]
c
in nearby neurons. The
[Ca
2⫹
]
c
signals are dependent on calcium influx from the extra-
cellular space and are inhibited by Zn
2⫹
or by the heavy-metal
chelator clioquinol. Our data are most readily consistent with a
model in which

A inserts into the plasma membrane, in which it
either forms channels or influences the properties of an existent
Ca
2⫹
-permeant channel. Several features of the responses are
remarkable in this respect: most notably, the selectivity of the
response for the astrocytes and the oscillatory, transient nature of
the [Ca
2⫹
]
c
signals. One might anticipate that insertion of Ca
2⫹
-
permeant channels into a cell membrane would generate a mono-
tonic increase in [Ca
2⫹
]
c
, and the appearance of the transient
fluctuations were surprising. However, the Mn
2⫹
quench and
confocal imaging data strongly argue that the transients do in-
deed result from transient episodic Ca
2⫹
influx and so presum-
ably reflect either the sporadic openings of a channel with a low
opening probability or the transient formation of channels that
then dissociate.
The other major original findings reported here are that

A
affects [Ca
2⫹
]
c
signals in astrocytes but not in neurons and that
Figure 6. Responses to

A are blocked by Zn
2⫹
and clioquinol but not by Cu
2⫹
.A, Addition of zinc (1 mMZnCl
2
) suppressed
the astrocyte response to

A. B, Addition of CuCl
2
(100
M) had no apparent effect on the

A responses, but the heavy-metal
chelator clioquinol (2
M)(C) suppressed the responses completely.
Figure 7.

A causes Ca
2⫹
-dependent depletion of GSH in both neurons and astrocytes.
MCBwasused to image astrocyteandneuronal GSH by digitalimaging.Hippocampal cocultures
(15–20 DIV) (A, B) and cortical astrocyte cultures ( C) were treated for 24 hr with

A25–35 or

A35–25 (50
Mfor both) and clioquinol (1
M) at 37°C in culture medium with (gray col-
umns) or without (black columns) calcium. Mean intensities of MCB–GSH adduct fluorescence
(arb.U) are presented.

A25–35 decreased GSH dramatically either in hippocampal astrocytes
in coculture with neurons ( A) or in cortical astrocytes in monoculture (C). The response was
dependent on extracellular calcium and was also suppressed by clioquinol ( C).

A35–25 had
no effect. Note that neuronal GSH was also significantly reduced ( B) and that the reduction was
also calcium dependent, although it represents a proportionately smaller response than that of
the astrocytes.
Abramov et al. •Astrocytic Calcium and GSH Changes in Amyloid Neurotoxicity J. Neurosci., June 15, 2003 •23(12):5088 –5095 • 5093
the ensuing neurotoxicity appears to be secondary to impaired
astrocytic function in the support of neuronal viability, although
we cannot exclude a direct toxic effect of the

A on neurons. In
most published studies of

A neurotoxicity, either neuronal cell
lines were used (Blanchard et al., 1997), or, in studies of neurons
in primary culture, the presence and contribution of glial cells
was not excluded (Mattson et al., 1992; Pike et al., 1993). The
selectivity of the effects on [Ca
2⫹
]
c
for astrocytes is remarkable
and may reflect some difference in the plasma membrane lipid
composition of the two cell types, because even small differences
in the lipid environment affect

A binding to membranes and
pore formation (Curtain et al., 2003). The stability of

A aggre-
gates in membranes is very delicately balanced (McLaurin and
Chakrabarty, 1996). For example, increased membrane choles-
terol, such as is found in AD brain and aging, favors a

-sheet
over an
␣
-helix conformation of

A (Curtain et al., 2003). To our
knowledge, very little information is available on the membrane
composition in different brain-cell types. Alternatively, the selec-
tivity may reflect the selective effects of

A on an existing chan-
nel, which is expressed in astrocytes but not in neurons, although
our pharmacological search has failed to reveal such a process.
In conclusion, our experimental conditions have allowed us to
uncover a novel toxic mechanism of

A, probably overlapping in
vivo with other known effects. This involves

-aggregation of

A
and the selective insertion into the astrocyte plasma membrane,
initiating sporadic [Ca
2⫹
]
c
signals, which can persist over long
periods. These signals, although not causing astrocyte cell death,
nevertheless promote GSH depletion in both cell populations;
they ultimately impair neuronal viability because GSH depletion
leaves the neurons vulnerable to damage by oxidative stress.
Thus, the resulting neurotoxicity reflects the central role of astro-
cytes in supporting neuronal function by supplying GSH precur-
sors and other metabolic intermediates and by removing excess
glutamate from the extracellular medium.
References
Arispe N, Rojas E, Pollard HB (1993) Alzheimer disease amyloid

protein
forms calcium channels in bilayer membranes: blockade by
tromethamine and aluminum. Proc Natl Acad Sci USA 90:567–571.
Arispe N, Pollard HB, Rojas E (1996) Zn
2⫹
interaction with Alzheimer
amyloid

protein calcium channels. Proc Natl Acad Sci USA
93:1710–1715.
Atwood CS, Moir RD, Huang X, Scarpa RC, Bacarra NM, Romano DM,
Hartshorn MA, Tanzi RE, Bush AI (1998) Dramatic aggregation of Alz-
heimer A

by Cu(II) is induced by conditions representing physiological
acidosis. J Biol Chem 273:12817–12826.
Behl C, Davis JB, Lesley R, Schubert D (1994) Hydrogen peroxide mediates
amyloid

protein toxicity. Cell 77:817–822.
Blanchard BJ, Konopka G, Russell M, Ingram VM (1997) Mechanism and
prevention of neurotoxicity caused by

-amyloid peptides: relation to
Alzheimer’s disease. Brain Res 776:40–50.
Boitier E, Rea R, Duchen MR (1999) Mitochondria exert a negative feed-
back on the propagation of intracellular Ca
2⫹
waves in rat cortical astro-
cytes. J Cell Biol 145:795–808.
Casley C, Land J, Sharpe M, Clark J, Duchen M, Canevari L (2002a)

-Amyloid fragment 25–35 causes mitochondrial dysfunction in primary
cortical neurons. Neurobiol Dis 10:258–267.
Casley CS, Canevari L, Land JM, Clark JB, Sharpe MA (2002b)

-Amyloid
inhibits integrated mitochondrial respiration and key enzyme activities.
J Neurochem 80:91–100.
Charles AC, Merrill JE, Dirksen ER, Sanderson MJ (1991) Intercellular sig-
naling in glial cells: calcium waves and oscillations in response to mechan-
ical stimulation and glutamate. Neuron 6:983–992.
Cherny RA, Atwood CS, Xilinas ME, Gray DN, Jones WD, McLean CA,
Barnham KJ, Volitakis I, Fraser FW, Kim Y, Huang X, Goldstein LE, Moir
RD, Lim JT, Beyreuther K, Zheng H, Tanzi RE, Masters CL, Bush AI
(2001) Treatment with a copper-zinc chelator markedly and rapidly in-
hibits beta-amyloid accumulation in Alzheimer’s disease transgenic mice.
Neuron 30:665–676.
Cuajungco MP, Goldstein LE, Nunomura A, Smith MA, Lim JT, Atwood CS,
Huang X, Farrag YW, Perry G, Bush AI (2000) Evidence that the

-amyloid plaques of Alzheimer’s disease represent the redox-silencing
and entombment of a

by zinc. J Biol Chem 275:19439–19442.
Curtain CC, Ali FE, Smith DG, Bush AI, Masters CL, Barnham KJ (2003)
Metal ions, pH and cholesterol regulate the interactions of Alzheimer’s
disease amyloid-

peptide with membrane lipid. J Biol Chem
278:2977–2982.
Dringen R (2000) Metabolism and function of glutathione in brain. Prog
Neurobiol 62:649–671.
Grynkiewicz G, Poenie M, Tsien RY (1985) A new generation of Ca
2⫹
indi-
cators with greatly improved fluorescence properties. J Biol Chem
260:3440–3450.
Harkany T, Abraham I, Timmerman W, Laskay G, Toth B, Sasvari M, Konya
C, Sebens JB, Korf J, Nyakas C, Zarandi M, Soos K, Penke B, Luiten PG
(2000) beta-amyloid neurotoxicity is mediated by a glutamate-triggered
excitotoxic cascade in rat nucleus basalis. Eur J Neurosci 12:2735–2745.
Haydon PG (2001) Glia: listening and talking to the synapse. Nat Rev Neu-
rosci 2:185–193.
He LM, Chen LY, Lou XL, Qu AL, Zhou Z, Xu T (2002) Evaluation of

-amyloid peptide 25–35 on calcium homeostasis in cultured rat dorsal
root ganglion neurons. Brain Res 939:65–75.
Kawahara M, Arispe N, Kuroda Y, Rojas E (1997) Alzheimer’s disease amy-
loid

-protein forms Zn
2⫹
-sensitive, cation-selective channels across ex-
cised membrane patches from hypothalamic neurons. Biophys J
73:67–75.
Keelan J, Allen NJ, Antcliffe D, Pal S, Duchen MR (2001) Quantitative im-
aging of glutathione in hippocampal neurons and glia in culture using
monochlorobimane. J Neurosci Res 66:873–884.
Kourie JI, Henry CL, Farrelly P (2001) Diversity of amyloid

protein frag-
ment [1–40]-formed channels. Cell Mol Neurobiol 21:255–284.
Lin H, Bhatia R, Lal R (2001) Amyloid

protein forms ion channels: impli-
cations for Alzheimer’s disease pathophysiology. FASEB J 15:2433–2444.
Figure 8.

A causes Ca
2⫹
-dependent cell death in neurons and not in astrocytes. Effect of

A on viability of neurons and astrocytes. PI fluorescence was used to detect dead cells 24 hr
after the addition of 50
M

A and 1
Mclioquinol in the presence or absence of Ca
2⫹
. Dead
cellswerecounted with respect to thetotalnumber of cells present,identifiedbystaining nuclei
with Hoechst 33342.

A caused a dramatic increase in cell death in neurons and only a modest
increase in astrocyte cell death in hippocampal cocultures. Cell death was calcium dependent
and cells were dramatically protected by 1
Mclioquinol.
5094 •J. Neurosci., June 15, 2003 •23(12):5088 –5095 Abramov et al. •Astrocytic Calcium and GSH Changes in Amyloid Neurotoxicity
Markesbery WR (1997) Oxidative stress hypothesis in Alzheimer’s disease.
Free Rad Biol Med 23:134–147.
Mattson MP, Chan SL (2001) Dysregulation of cellular calcium homeostasis
in Alzheimer’s disease: bad genes and bad habits. J Mol Neurosci
17:205–224.
Mattson MP, Cheng B, Davis D, Bryant K, Lieberburg I, Rydel RE (1992)

-Amyloid peptides destabilize calcium homeostasis and render human
cortical neurons vulnerable to excitotoxicity. J Neurosci 12:376–389.
Matyash M, Matyash V, Nolte C, Sorrentino V, Kettermann H (2002) Re-
quirement of functional ryanodine receptor type 3 for astrocyte migra-
tion. FASEB J 16:84–86.
McLaurin J, Chakrabartty A (1996) Membrane disruption by Alzheimer

-amyloid peptides mediated through specific binding to either phos-
pholipids or gangliosides: implications for neurotoxicity. J Biol Chem
271:26482–26489.
Meske V, Hamker U, Albert F, Ohm TG (1998) The effects of

/A4-amyloid
and its fragments on calcium homeostasis, glial fibrillary acidic protein
and S100

staining, morphology and survival of cultured hippocampal
astrocytes. Neuroscience 85:1151–1160.
Melov S (2002) “. . . and C is for Clioquinol”: the A

Cs of Alzheimer’s dis-
ease. Trends Neurosci 25:121–123.
Muller WE, Romero FJ, Perovic S, Pergande G, Pialoglou P (1997) Protec-
tion of flupirtine on

-amyloid-induced apoptosis in neuronal cells in
vitro: prevention of amyloid-induced glutathione depletion. J Neuro-
chem 68:2371–2377.
Peuchen S, Clark JB, Duchen MR (1996) Mechanisms of intracellular cal-
cium regulation in adult astrocytes. Neuroscience 71:871–883.
Pike CJ, Burdick D, Walencewicz AJ, Glabe CG, Cotman CW (1993) Neu-
rodegeneration induced by

-A peptides in vitro: the role of peptide
assembly state. J Neurosci 13:1676–1687.
Pizzo P, Burgo A, Pozzan T, Fasolato C (2001) Role of capacitative calcium
entry on glutamate-induced calcium influx in type-I rat cortical astro-
cytes. J Neurochem 79:98–109.
Rovira C, Arbez N, Mariani J (2002) A

(25–35) and A

(1–40) act on
different calcium channels in CA1 hippocampal neurons. Biochem Bio-
phys Res Commun 296:1317–1321.
Sagara J, Miura K, Bannai S (1993) Maintenance of neuronal glutathione by
glial cells. J Neurochem 61:1672–1676.
Schubert P, Ogata T, Marchini C, Ferroni S (2001) Glia-related patho-
mechanisms in Alzheimer’s disease: a therapeutic target? Mech Ageing
Dev 123:47–57.
Small DH, McLean C (1999) Alzheimer’s disease and the amyloid-

pro-
tein: what is the role of amyloid? J Neurochem 73:443–449.
Stix B, Reiser G (1998)

-Amyloid peptide 25–35 regulates basal and
hormone-stimulated Ca
2⫹
levels in cultured rat astrocytes. Neurosci Lett
243:121–124.
Takahashi M, Billups B, Rossi D, Sarantis M, Hamann M, Attwell D (1997)
The role of glutamate transporters in glutamate homeostasis in the brain.
J Exp Biol 200:401–409.
Ueda K, Shinohara S, Yagami T, Asakura K, Kawasaki K (1997) Amyloid

protein potentiates Ca
2⫹
influx through L-type voltage-sensitive Ca
2⫹
channels: a possible involvement of free radicals. J Neurochem
68:265–271.
Varadarajan S, Yatin S, Aksenova M, Butterfield DA (2000) Alzheimer’s
amyloid

-peptide-associated free radical oxidative stress and neurotox-
icity. J Struct Biol 130:184–208.
Vergun O, Sobolevsky AI, Yelshansky MV, Keelan J, Khodorov BI, Duchen
MR (2001) Exploration of the role of reactive oxygen species in gluta-
mate neurotoxicity in rat hippocampal neurones in culture. J Physiol
(Lond) 531:147–163.
White AR, Bush AI, Beyreuther K, Masters CL, Cappai R (1999) Exacerba-
tion of copper toxicity in primary neuronal cultures depleted of cellular
glutathione. J Neurochem 72:2092–2098.
Abramov et al. •Astrocytic Calcium and GSH Changes in Amyloid Neurotoxicity J. Neurosci., June 15, 2003 •23(12):5088 –5095 • 5095